Sales teams argue about follow-up counts the way developers argue about tabs versus spaces. Six touches. Nine. Twelve. The number is the wrong debate. The right question is what each follow-up is actually for, and how the sequence knows when to shut up.
This piece names a specific framework we use at Milo and ship as a product feature: the stop-on-reply cadence. Four messages, four intents, hard stop on any reply. It is opinionated on purpose. You can lengthen it, but you cannot skip the stop rule.
Why do most follow-up sequences leak?
Backoff.ai and similar deliverability studies keep landing on the same finding: single-touch outbound gets a small fraction of the responses that a proper cadence gets. Woodpecker looked at millions of sent messages and found reply rates climb materially between message one and message four, then plateau. The upside is real, but only if the last three messages are not clones of the first.
Woodpecker cold email response rate research Cold email response rate research.
The other leak is worse. Sequences that keep sending after a reply. A prospect writes back with a real question, and 48 hours later your tool sends the "did you see my last note" bump. You look like a bot. You are a bot. Every deliverability guide from Google Postmaster to the M3AAWG sender BCP treats "recipient signals ignored" as a reputation issue, not a copy issue.
M3AAWG Sender Best Common Practices M3AAWG Sender BCP (PDF).
The stop-on-reply cadence, in one picture
Ladder
The stop-on-reply cadence
- 01
Day 0: Signal-anchored open
Name the signal you saw. One ask.
- 02
Day 3: Short nudge
Reply to your own thread. Two lines.
- 03
Day 7: Value-add
Send something useful. No ask.
- 04
Day 14: Last call
Explicit close. Permission to end.
- 05
Reply detected -> cadence stops
Any human reply, at any step, ends the sequence.
The cadence table
| # | Day | Intent | Template shape |
|---|---|---|---|
| 1 | Day 0 | Signal open | One sentence naming the signal, one sentence with your angle, one question. Under 90 words. |
| 2 | Day 3 | Short nudge in-thread | Reply to your own message. Two lines. Restate the ask in different words. No new attachments. |
| 3 | Day 7 | Value-add, no ask | Send a specific artifact: a checklist, a competitor teardown, a link to a fix. The email closes without a CTA. |
| 4 | Day 14 | Last call with permission to end | Explicit close. "If this is not a fit right now, I will stop here. Otherwise, one line and I will send a Cal link." |
- #
- 1
- Day
- Day 0
- Intent
- Signal open
- Template shape
- One sentence naming the signal, one sentence with your angle, one question. Under 90 words.
- #
- 2
- Day
- Day 3
- Intent
- Short nudge in-thread
- Template shape
- Reply to your own message. Two lines. Restate the ask in different words. No new attachments.
- #
- 3
- Day
- Day 7
- Intent
- Value-add, no ask
- Template shape
- Send a specific artifact: a checklist, a competitor teardown, a link to a fix. The email closes without a CTA.
- #
- 4
- Day
- Day 14
- Intent
- Last call with permission to end
- Template shape
- Explicit close. "If this is not a fit right now, I will stop here. Otherwise, one line and I will send a Cal link."
Message 1: the signal open (Day 0)
The first email decides whether the follow-ups ever get opened. If it does not name a real signal, follow-ups will not save it. A signal is something concrete you observed: a permit filed, a job posting live, a page missing schema, a review that just landed, a service page with no phone number. If your first line reads like it could go to anyone in the industry, rewrite it.
Template shape (Day 0)
Subject: quick note on {specific signal}. Body: I saw {signal, one sentence with source}. Most {vertical} folks fix this by {one-sentence angle}. Worth a 12 minute call this week to walk you through what we did for {peer}?
Message 2: the two-line nudge (Day 3)
Reply to your own thread. Not a new subject line. Two lines maximum. If message 1 was crisp, this one is almost boring, and that is the point. You are lifting the thread back to the top of the inbox without adding cognitive load.
Template shape (Day 3)
Bumping this in case it slipped. Same ask as above: 12 minutes this week to show what we did for {peer}?
Message 3: the value-add with no ask (Day 7)
This is where most cadences quietly die. The third touch is another version of "just checking in" and it deserves to be ignored. Instead, send something the prospect can use even if they never talk to you. A one-pager. A tear-down of a competitor page. A specific fix for the signal you named on day zero. Then close the email. No call to action. This message earns the fourth.
Template shape (Day 7)
Sharing a short teardown of {peer} in case it is useful even if we never speak. Two things they got right, one thing you could beat them on: {link or bullets}. No ask, just wanted to send this over.
Message 4: the last call (Day 14)
Explicitly give the prospect permission to end the conversation. This one line does more work than any subject line trick: "if this is not a fit right now, I will stop here." It also protects your reputation. You told them you would stop. Stop.
Template shape (Day 14)
Last note from me on this. If this is not a fit right now, no problem, I will stop here. If it is, reply with any word and I will send a Cal link.
The one thing to remember
A cadence works when each rung carries a different intent and the whole sequence stops the instant a human reply lands. Reply-then-bump is the single fastest way to burn a domain, because postmaster tools read ignored recipient signals as complaints, not copy problems. The stop rule is worth exactly as much as the tool that enforces it.
Google email sender guidelines Google sender guidelines.
Why 0, 3, 7, and 14 for day gaps?
The gaps are not arbitrary but they are also not sacred. The principle is that each gap doubles roughly, which mirrors how humans forget and re-remember. Day 3 catches the "meant to reply, got busy" cohort. Day 7 catches "was traveling that week." Day 14 catches "budget cycle just opened." Beyond 14 days, response probability collapses and every additional touch trades reply rate for spam risk.
If your industry is slow (procurement, education, public sector), stretch to 0, 5, 12, 21. If it is fast (agencies, ecommerce ops, local services), compress to 0, 2, 5, 10. Keep four rungs. Keep the stop rule.
What is each message not?
- Message 2 is not a repeat of message 1 with a different subject line. Same thread, two lines, different words.
- Message 3 is not "just floating this to the top." It is a useful artifact with no ask. If you have nothing useful, skip to message 4.
- Message 4 is not "one last time before I close the file." It is explicit permission to end the conversation.
- None of the messages contain "did you get my last email." That phrase is a tell.
Which metrics should you watch?
The cadence is working when reply rate climbs from message 1 to message 3 and the total sequence complaint rate stays under Google's 0.10 percent hard threshold. If message 4 is producing your highest complaint rate, the last-call line is too aggressive; soften it. If message 3 is producing near-zero replies, your value-add is not specific enough.
Google Postmaster spam rate guidance Postmaster spam rate thresholds.
FAQ
Can I go past four messages?
You can. Reply rates plateau after touch four in most studies, and complaint rates start climbing. If you extend, add rungs that carry new intent (a case study, a peer comparison), not another "checking in."
What counts as a "reply" for the stop rule?
Any human-authored inbound on the thread. Out-of-office autoresponders do not count; the sequence pauses until the OOO end date, then resumes. Anything from a real person, including "not interested," stops the cadence for good.
Should follow-ups be in the same thread or a new one?
Same thread for messages 2, 3, and 4. It keeps the conversation coherent for the prospect and lowers the chance a follow-up looks like a fresh cold email to spam filters.
Does Milo enforce the stop rule automatically?
Yes. Reply detection is tied to sequence state; queued follow-ups are cancelled the moment an inbound reply is logged on the thread by Gmail or Outlook. It is not a manual toggle.
How do I know if my Day 0 signal is strong enough?
Read the first sentence out loud. If it could plausibly have been sent to a hundred other companies with a search-and-replace, the signal is too weak.